Assessment of existing concrete bridge girder for reuse based on acoustic emission data from load testing

Abstract

Cement production is a major contributor to global warming, responsible for around 6\% of annual greenhouse gas emissions. Rijkswaterstaat aims to operate fully circularly by 2030, which makes high-value reuse of structural elements an attractive strategy to reduce CO\textsubscript{2} without compromising safety. This thesis investigates whether non-destructive proof-load testing with acoustic emission (AE) can provide sufficient evidence to justify circular reuse of prestressed concrete girders that fail current code checks on paper.

The work develops and applies a framework that combines code-based resistance models from Eurocode 2:2004 and the next-generation Eurocode 2:2023, a controlled proof-load protocol with AE monitoring, and a probabilistic reliability analysis. The case study concerns reclaimed inverted T-girders without effective shear reinforcement. At the shear control section, the governing ultimate shear effect is \(V_{ULS}=613.13\ \mathrm{kN}\), while code checks give \(V_{Rd,c}=417.11\ \mathrm{kN}\) (EC2:2004) and \(V_{Rd,c}=357.37\ \mathrm{kN}\) (EC2:2023). On this basis alone, reuse at consequence class 3 over 100 years would be rejected.

In the test, one girder was proof-loaded to 550 kN without irreversible damage under the green light criteria and then taken to failure for research purposes. AE processing, using a data-driven peak-frequency threshold and a calibrated wave speed, revealed first flexural activity around 462 kN and first shear-type activity around 589 kN, ahead of visible cracks. The girder ultimately failed at an applied load of 1046 kN, corresponding to a sectional shear of about 875.1 kN at the control section, highlighting a significant gap between sectional code predictions and the observed global capacity.

AE also enabled back-calculation of the effective prestress from the first flexural cracking, refining it from 2507 kN to 2686 kN. Reliability was quantified using Monte Carlo simulation over a 100-year horizon. Without updating, the 100-year reliability indices are \(\beta_{100}=2.81\) (EC2:2004) and \(\beta_{100}=2.76\) (EC2:2023), well below the CC3 target of 4.27. Conditioning on survival at 550 kN and incorporating the AE-informed prestress update raises the reliability index only to \(\beta_{100}=2.98\ (+0.17)\) (EC2:2004) and \(\beta_{100}=2.88\ (+0.12)\) (EC2:2023), still below the target.

The main conclusion is that AE-assisted proof-load testing substantially improves observability and strengthens the evidence for reliability assessment, but for this girder type, probabilistic reliability model, and load model it does not, by itself, achieve CC3-level reliability at full capacity. For circular reuse, a different resistance model in probabilistic framework or additional measures are required, such as demand reduction in the receiving structure, selective strengthening, or refined resistance modelling calibrated to a broader test set. Future work should focus on validating AE with digital image correlation, optimizing proof-load resistance based on traffic light criterion, optimizing instrumentation, and quantifying model discrepancy between code predictions and observed behaviour.